Goals and Deliverables

1. Xrootd infrastructure:
2. Fallback usage: Ability for a reasonably small subset of jobs at a T2 site to run on remote data via the global infrastructure. These jobs may use the fallback when the local copy of a file is unexpectedly able to open, or when the job is purposefully sent to a site without the file ("overflow").
3. Interactive Use: A user can easily and efficiently use the global infrastructure for the event display, interactive ROOT/FWLite sessions, or single cmsRun jobs based on their personal computer.
4. Diskless T3 usage: Ability for a small (8-20 node) T3 site to be fully utilized by local jobs (analysis or production) using the global infrastructure as its only means of data access.

Timeline

This timeline is organized into four checkpoints, taking us from beginning February 2011 through August 2011. Each checkpoint has several milestones and may have a deliverable that accomplishes the high-level goals above.

• March 1:
  • Have basic monitoring available: Service heartbeat tests, email alerts, ML-based status monitoring (connected clients, open files, MB/s)
  • CMS code requirements clearly understood and agreed upon by the xrootd team.
    • Particularly, ability to throttle namespace queries should be delivered here.
  • Basic software packaging and documentation for sites.
  • DELIVERABLE: Have a minimal production and integration Xrootd infastructure for sites to join.
• May 1:
  • JobRobot-equivalent test for Xrootd.
  • Integrate basic service tests into an appropriate larger-scale framework (such as RSV or SAM) instead of the parochial central one used for March 1.
  • Improve monitoring, allowing sysadmins to be able to identify and alert on abusive users.
  • DELIVERABLE: Fallback usage case.
• June 1:
  • DELIVERABLE: Overflow fallback usage case.
• August 1:
  • Note: all sites should have had three months to stabilize. Effort for developing new things ought to be reduced as we must maintain more services.
  • DELIVERABLE: Interactive use.
    • Note we will not actively prevent users from using the production infrastructure up to this point; we just will not encourage it.
  • DELIVERABLE: Diskless T3 usage.
    • Note we will not actively prevent T3 sites from using the production infrastructure up to this point; we just will not encourage it.

Note we have left site penetration goals off the timeline. This is currently on purpose (may change) - the timeline is focusing on getting ready for sites to participate.